Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Capital Markets Review Vol. 24, No.1, pp. 16-37 (2016) ISSN 1823-4445
16
Effect of Geographical Diversification on Informational
Efficiency in Emerging Countries
Suan Poh* & Chee-Wooi Hooy
School of Management, Universiti Sains Malaysia
Abstract: This study investigates the effect of geographical diversification on
informational efficiency. Informational efficiency is measured using the price delay
measure and is further divided into informational efficiency related to local news and
informational efficiency related to global news. Geographical diversification is
proxied using four different variables – foreign sales dummy, number of foreign
countries, foreign sales ratio and the Herfindahl Index. The sample study involves
public listed companies from 12 emerging countries for the period from year 2005 to
year 2014. The regression results prove that all four geographical diversification
proxies show a positive and significant effect on local price delay. This proves that
when a company undergoes geographical diversification, its business structure will
become complex. Few investors will focus on a geographical diversified company,
which leads to lower informational efficiency. All the regression models in this study
are robust to heteroscedasticity and multicollinearity problems. All the geographical
diversification proxies remain significant towards local delay when alternative delay
measures are used, and financial crisis is controlled by using a crisis dummy.
Keywords: Informational efficiency, Price delay, Geographical diversifications,
Emerging markets
JEL Classification: F23, G14, G15
1. Introduction
Over the past 25 years, increasingly integrated capital markets and globalization have
lowered the cost of companies doing business in foreign markets. Foreign investment made
by corporations in the industrialized nations has grown dramatically. Generally, firms adopt
geographical diversification with similar business operations in different countries as the
main corporate strategy to gain competitive advantages (Barney and Hesterly 2008; Chang
and Wang 2007; Hitt et al. 1997). For example, large publicly traded US and EU firms
operate their businesses, on average, in more than three different geographic markets
(Bodnar et al. 1999; Pavelin and Barry 2005).
Geographical diversification is said to confer a number of advantages, including full
use of resources and distribution of costs on the basis of the growing market and product
range, which lead to economies of scale (Ghoshal 1987). To the extent that firms are able to
leverage their operations worldwide, international investment may enable them to capture
valuable operating synergies (Feinberg and Phillips, 2002). Human capital in multinational
companies can learn and innovate faster (Bartlett and Ghoshal 2000). Companies that are
able to expand their business globally can gain access to specific skills at lower cost (Kogut
1985; Porter 1986). Companies can also shift their production lines to other countries with
lower labour costs (Kogut and Kulatilaka 1994).
This paper departs from the traditional focus of geographical diversification on benefits
and cost of firm values to a relatively less explored area, the informational efficiency of
stock markets. Market efficiency can be defined as the extent and speed that market prices
of tradable assets incorporate the available information. High market efficiency means that
the process of incorporating information into market prices is fast and complete.
* Corresponding author: Suan Poh. Email: [email protected]
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
17
Since Professor Eugene Fama introduced his Efficient Market Hypothesis (EMH),
“informational efficiency” has become a common term in financial studies. Informational
efficiency is used to indicate the extent that market prices of tradable assets incorporate the
available information. After the theory of EMH was introduced, researchers started to test
the EMH theory in the stock markets around the world. In the early part of the research,
researchers only focused on testing whether or not a market was efficient. They used many
different methods to test EMH, such as serial correlation tests (Fama 1965), spectral
analysis as used by Granger and Morgenstern (1963), and the variance ratio test (Lo and
MacKinlay 1988).
In 1997, Campbell et al. (1997) offered the concept of “relative efficiency”, which is
the informational efficiency of a market measured relative to another market. Not only can
we indicate whether a market is efficient or not, but also the lead-lag relationship between
any two markets can be determined. However, it is still not possible to quantify the degree
of efficiency of a market. Hou and Moskowitz (2005) proposed a price delay model that can
quantify the informational efficiency of a market. This price delay model has several
advantages compared to conventional tests. First, it permits researchers to identify various
factors that affect the informational efficiency of stocks. Second, it enables researchers to
measure the adjustment of stock price to both local and global market information.
Since this price delay measure has been proposed, many studies have been done to
investigate the determinants of informational efficiency. Basically, the determinants of
informational efficiency can be divided into two sections – stock specification and firm
fundamental. Considerable research has been done on variables about stock specifications,
for example, trading volume by Chordia and Swaminathan (2004), the liberalization process
by Bae et al. (2012), liquidity by Lesmond (2005), short selling by Saffi and Sigurdsson
(2011), option pricing by Phillips (2011), and market frictions by Hou and Moskowitz
(2005). However, only a few researchers have studied the relationship between firm
fundamentals and informational efficiency, such as firm size by Hou and Moskowitz (2005),
accounting quality by Callen et al. (2013) and analyst coverage by Bae et al. (2012). The
price delay model is still incomplete, as there are still many other undiscovered factors that
affect the informational efficiency of a firm.
This study focuses on public listed companies from emerging markets to study the
effect of geographical diversification on informational efficiency. Emerging countries
present an interesting case study for geographical diversification because there have been a
huge increase in the outflows of foreign direct investment (OFDI) from emerging markets in
the 2000s.
OFDI streams from emerging market multinational enterprises (MNEs), which are
indicated in Table 1 by companies from both developing countries and transition economies,
have demonstrated especially dynamic growth rates of roughly 300% from US$159 billion
in 2005 (US$140 billion from developing countries and US$19 billion from transition
economies), to reach approximately US$482 billion in 2012 (US$426 billion from
developing countries and US$55 billion from transition economies).
In recent decades, the OFDI of emerging countries has changed tremendously in terms
of regional distribution. Emerging market MNEs have increased their foreign investment in
many other developing countries. They have progressively increased the resources
allocation in developed countries. Among all the sectors, firms that export natural resources
and service firms make up the highest percentage of OFDI (World Investment Report 2008).
In this study, the relationship between geographical diversification and informational
efficiency is investigated. We further divide informational efficiency into local information
and global information, which is discussed separately in hypotheses 1 and 2. Each
hypothesis is further divided into four subsections that represent four different proxies for
Suan Poh and Chee-Wooi Hooy
18
geographical diversification – foreign sales dummy, number of foreign countries, foreign
sales ratio and the Herfindahl Index – in order to capture the different dimensions of
geographical diversification. We postulate that the different dimensions of geographical
diversification have a different effect on informational efficiency.
Table 1: FDI outflows, by region and economy
Region/economy USD‟ millions
2005 2006 2007 2008 2009
World 903,763 1,427,473 2,272,048 2,005,332 1,149,776
Developed economies 744,407 1 152196 1,890,419 1,600,707 828,005
Developing economies 139,934 244,703 330,033 344,033 273,401
Transition economies 19,422 30,573 51,596 60,591 48,368
Region/economy 2010 2011 2012 2013 2014
World 1,504,927 1 678 035 1 390 956 1,305,910 1,354,046
Developed economies 1,029,836 1 183088 909,383 833,630 822,826
Developing economies 413,219 422,066 426,081 380,784 468,184
Transition economies 61,871 72,879 55,491 91,496 63,072
Source: UNCTAD, World Investment Report 2015. (http://unctad.org/en/PublicationsLibrary/wir2015_en.pdf)
Foreign sales dummy is able to separate diversified firms and focused firms into two
groups and compare their influence on informational efficiency related to local news.
According to Chen (2005), individual investors and institutional investors prefer
information that is easy to understand and widely available. When a company undergoes
geographical diversification, its business coverage area becomes larger and the company is
exposed to other countries‟ risk where its business is involved. Its business structure
becomes more complex than that of a company that only focuses its sales locally (Morck
and Yeung 1991). As a result, they will not pay attention to a firm that has diversified in
many different foreign countries, which, consequently, will cause the potential investor base
of the diversified firm to become smaller. Eventually the informational efficiency will
become lower.
Hypothesis 1(a): Foreign sales dummy has a negative and significant effect on
informational efficiency related to local news.
Number of foreign countries captures the width of geographical diversification. In this
study, we suggest that the number of foreign countries will have the same result as the
foreign sales dummy, which has a negative effect on informational efficiency.
Hypothesis 1(b): Number of foreign countries has a negative and significant effect on
informational efficiency related to local news.
Foreign sales ratio shows the percentage of the total sales of a company that comes
from foreign countries. Foreign sales ratio measures the depth of geographical
diversification. We suggest that foreign sales ratio will have the same result as the foreign
sales dummy, which has a negative effect on informational efficiency.
Hypothesis 1(c): Foreign sales ratio has a negative and significant effect on informational
efficiency related to local news.
The Herfindahl Index is used to capture both the width and intensity of the geographical
diversification (Hitt et al. 1997; Denis et al. 2002). We find that geographical diversification
has two effects on the investor base of a firm, which will significantly affect its
informational efficiency. Firstly, the investor base of a company will increase by the
inclusion of foreign investors that recognize the company through its products or business in
a foreign country, which will subsequently increase its informational efficiency. Secondly,
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
19
geographical diversification will cause a company‟s structure to become more complex.
This will cause fewer investors to pay attention to it since most of the investors prefer a
simple company that is easy to analyse. We suggest that the complexity effect will
outperform the foreign investors‟ recognition effect and Herfindahl Index will have a
negative effect on informational efficiency.
Hypothesis 1(d): Herfindahl Index has a negative and significant effect on informational
efficiency related to local news.
For Hypothesis 2, we investigate the effect of geographical diversification on the
informational efficiency related to global news instead of local news. First, we use the
foreign sales dummy as a proxy. Foreign sales dummy is able to separate diversified firms
and focused firms into two groups and compare their influence on informational efficiency
related to global news. Merton (1987), in his investor recognition theory, mentions that
investors only give consideration to a limited number of stocks and only exchange stocks
that they have information about. Consequently, stocks that are less known by investors
have a smaller potential speculator base. Because of the restricted consideration, speculators
can only consider a subset of all the accessible data. They generally do not consider or focus
on data from stocks that they do not take part in. As a result, if a company has sales in
foreign countries, it will be able to attract more attention from foreign investors and
eventually increase its informational efficiency. Lim and Hooy (2013) also used foreign
sales dummy as a moderator in the relationship between foreign shareholdings and
informational efficiency. They mentioned that diversified firms can increase their visibility
to foreign investors by diversifying their business to foreign countries. Their products will
become more recognizable to Foreign investors who will invest in their stocks. In our
research, we agree with this statement and suggest that the foreign sales dummy will
positively affect a firm‟s informational efficiency.
Hypothesis 2(a): Foreign sales dummy has a positive and significant effect on informational
efficiency related to foreign news.
Number of foreign countries captures the width of geographical diversification. We
postulate that the more countries that a firm diversifies in, the more foreign investors will
recognize the company. As a result, we suggest that the number of foreign countries will
have the same result as the foreign sales dummy, which has a positive effect on
informational efficiency.
Hypothesis 2(b): Number of foreign countries has a positive and significant effect on
informational efficiency related to foreign news.
Foreign sales ratio shows the percentage of the total sales of a company that comes
from foreign countries, and measures the depth of geographical diversification. We
postulate that the more foreign sales a company has in foreign countries, the larger the
potential foreign investor base it has. Therefore, we suggest that the foreign sales ratio will
have the same result as the foreign sales dummy, which has a positive effect on
informational efficiency.
Hypothesis 2(c): Foreign sales ratio has a positive and significant effect on informational
efficiency related to foreign news.
Herfindahl Index captures both the width and intensity of the geographical
diversification (Hitt et al. 1997; Denis et al. 2002). The investor base of a company that has
a higher value of Herfindahl Index will increase by the inclusion of foreign investors who
recognize the company through its products or business in a foreign country. This will
subsequently increase its informational efficiency. We suggest that the Herfindahl Index
will have the same result as the number of foreign countries and foreign sales ratio since it is
a combination of both and will have a positive effect on informational efficiency.
Suan Poh and Chee-Wooi Hooy
20
Hypothesis 2(d): Herfindahl Index has a positive and significant effect on informational
efficiency related to foreign news.
The remainder of the paper is structured as follows. Section 2 discusses the variables
and model specification. Section 3 discusses the empirical results. Section 4 presents the
conclusion.
2. Methodology and Data
2.1 Measurement of Variables
In this subsection, the three main types of variable – dependent variable, independent
variable and control variable – are discussed and the methods used to compute each variable
are explained in detail.
2.1.1 Price Delay as Dependent Variable
Our construction of the local and global price delay measures follows the framework of Bae
et al. (2012), which involves the following unrestricted model:
(1)
where is the return on stock at week and denote the contemporaneous
and four weekly lagged returns on the local and world market indices, respectively. We
follow the convention in the price delay literature in utilizing weekly instead of monthly or
daily returns. As indicated by Hou and Moskowitz (2005), the dispersion for monthly data is
small because, generally, the information is incorporated into the stock price within one
month, whereas, for daily returns, there are numerous microstructure impacts that influence
the results, such as non-synchronous trading. The construction requires the following two
restricted models:
(2)
(3)
For each year, from 2005 through to 2014, we estimate equations (4) through (6) for every
firm in the sample. Their respective R-squares are used to calculate the scaled version of
stock price delay for firm i in year t:
(4)
(5)
( ) captures the variation in contemporaneous individual stock returns that
is explained by the lagged returns on the local (world) market index, where the latter is used
as a market-wide information signal. The greater the explanatory power of these lags, the
4 4
, , , , , ,
0 0
i t i i k m t k i k w t k i t
k k
r r r
,i tr i ,t ,m t kr ,w t kr
4
, 0 , , , ,
0
i t i i m t i k w t k i t
k
r r r
4
, , , 0 , ,
0
i t i i k m t k i w t i t
k
r r r
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
21
longer the delay in responding to market-wide news that has common effects across firms.
The value of is limited between zero and one, with a value closer to zero (one)
showing the faster (slower) incorporation of information, and hence a higher (lower) degree
of stock price efficiency. The data required for the calculation of these stock price delay
measures are the weekly closing prices for individual stocks, the local market index and the
world market index. Following the common practice, weekly returns are calculated by
compounding daily returns between adjacent Wednesdays in order to avoid market
anomalies, such as the weekend and Monday effects (Bartholdy and Peare 2005). Hou and
Moskowitz (2005) contended that lower frequency data like monthly data will lead to
estimation error. For monthly data, there is little dispersion in the price delay measure since
the stock prices respond to information within a month. On the other hand, although high
level frequency data, such as daily data, provide more precision and greater dispersion in
price delay, the daily data are also influenced by confounding microstructure problems, such
as nonsynchronous trading and bid-ask bounce.
2.1.2 Geographical Diversification as Independent Variable
Geographical diversification has typically been measured in terms of the intensity of
international involvement and the geographic scope of international operations, as
highlighted by Lu and Beamish (2004). This study employs several types of geographical
diversification proxies in order to capture different aspects of geographical diversification.
Four different methods are used to measure geographical diversification in this study.
The first indicator used is the foreign sales dummy variable (DIVERSE). Firms with a
foreign sales to total sales ratio of more than 10% are classified as diversified. Firms that do
not fulfil the conditions are classified as focused (John and Ozgur 2006).
The second indicator that is used is the number of foreign countries (FCOUNTRY).
This indicator shows the total number of foreign countries in which a company diversifies
(Tallman and Li 1996).
The third indicator used is foreign sales ratio (FSALES). All the sales recorded outside
the country in which the company is registered are perceived as foreign sales (Tallman and
Li 1996).
FSALES = Foreign Sales/ Total Sales (6)
The fourth indicator used is the Herfindahl Index (HERFINDAHL), which is
constructed from foreign sales in each foreign country; this is a measure that has been
commonly used in many previous studies examining diversification issues (Hitt et al. 1997;
Denis et al. 2002). The Herfindahl index is calculated as follows for each firm i:
HERFINDAHL =1- Σ(Sales per country/Total sales) (7)
The Herfindahl Index ranges from 0 to 1. The closer the Herfindahl Index is to 1, the
more a firm‟s sales are diversified geographically, and the closer it is to 0, the more the
firm‟s sales are concentrated in a few countries.
2.1.3 Control Variables
The literature review of informational efficiency shows that numerous researchers have
concentrated on finding the determinants of price delay since Hou and Moskowitz (2005)
introduced the price delay model to quantify informational efficiency. Thus, there are a few
variables that are normally used as control variables in the price delay model. The four
DELAY
Suan Poh and Chee-Wooi Hooy
22
control variables that are utilized in this study are firm size, trading volume,
liquidity/transaction costs and the number of security analyst's coverage.
First and foremost, firm size (lnMCAP) is measured using the natural logarithm of the
annual market capitalization of a company at the end of the calendar year. Previous
researchers, such as Lim and Hooy (2010), Phillips (2010), Saffi and Sigurdson (2011), and
Hou and Moskowitz (2005), incorporated firm size as a control variable in their study.
Secondly, since Chordia and Swaminathan (2000) demonstrated that high trading
volume stocks tend to be promptly changed in accordance with new market information
compared to low trading volume stocks, trading volume has turned into an imperative
determinant for price delay. For example, Bae et al. (2012), Callen et al. (2013), and Lim
and Hooy (2010), included this control variable in their study of informational efficiency. In
this study, trading volume (VOLUME) is proxied by the average monthly share trading
volume scaled by total shares outstanding for each firm and year.
Thirdly, the proxy for volatility is the standard deviation of weekly returns for a year
(VOLATILITY) (Bae et al. 2012). Thomson Reuters DataStream gives the firm-level panel
data of annual market capitalization, monthly share volume, and weekly closing stock
prices.
Fourthly, analyst coverage is one of the most vital control variables, and has been
incorporated by researchers such as Hou and Moskowitz (2005), and Bae et al. (2012) in
their price delay model. From the perspective of the data given by the Institutional Brokers
Estimate System (I/B/E/S), the number of analysts issuing earnings forecasts (ANALYST)
for a firm each year is collected. Analyst coverage is written as being equivalent to zero for
a firm-year observation if a firm is not listed on the I/B/E/S database or does not have
earnings forecasts for any given year.
2.2 Model Specification
In this study, we use multiple regression to test the hypotheses in this study. Multiple
regression is used when there are many independent variables but only one dependent
variable. The independent variables in this study include a geographical diversification
proxy, which acts as the subject variable and other prominent control variables from
previous literature. Figure 1 shows the empirical framework.
For the control variables, we include several important variables – firm size, trading
volume, liquidity and analyst forecasts. These variables have been proven to have a
significant effect on price delay in the previous literature pertaining to informational
efficiency. The OLS estimator is a method for estimating a well fitted regression line by
minimizing the residual sum of squares. OLS is appropriate in this study as it is the most
straightforward regression technique, and the estimation is reliable as long as common
regression problems are accounted for. The pooled ordinary least squares (OLS) regression
model is specified as follows:
( )
(8)
2.3 Data Analysis
The data are considered as panel data as they involve public listed firms in emerging
countries that span across 10 years. We choose to collect the data as unbalanced data as each
firm may have a varying number of observations. The reason for this is that in many
countries, geographical diversification data do not have to be disclosed in the annual report
but are left to the free will of each respective firm. Therefore, we predict that there will be
incomplete data across the period that we study. Our panel data represent a short panel as
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
23
Figure 1: Empirical framework
there are large number of firms with a short time period. The number of firm observations is
larger than the number of year observations.
2.3.1 Sample Data
Our sample includes public listed companies in emerging countries following the IMF list.
However, due to the unavailability of data for the sales data and analysts‟ forecasts
according to the geographic segment, some countries have been excluded from our sample
study. As a result, there are only 12 countries remaining, which are Argentina, Brazil, Chile,
China, India, Indonesia, Malaysia, Mexico, Philippines, South Africa, Thailand and Turkey.
To construct the price delay variables, which are the dependent variables in the study,
we require stock prices for each public listed company, as well as the local stock index for
each emerging country and the MSCI stock index. To construct geographical diversification
proxies, the foreign sales of each public listed company in the emerging countries and the
foreign countries that the company has diversified in are needed. To construct the control
variables, the market capital, stock volumes, stock daily returns and number of analysts that
study the company are required. Finally, for further study, the company components of each
local index and the number of industries in which the company is involved are needed.
2.3.2 Time Span
The time span of this sample study ranges from year 2005 to year 2014. Since our sample
data concentrate on public listed companies in emerging countries, during this time span, the
OFDI of emerging countries has increased by a considerable amount (World Investment
Report). The increase in OFDI shows that many companies in emerging countries are
actively involved in geographical diversification to foreign countries. The objective of this
study is to analyse the effect of geographical diversification on informational efficiency.
The considerable increase in OFDI will improve the significance of the results.
3. Results and Discussions
3.1 Descriptive Statistics
There are three tables for the overall descriptive statistics. Table 2 describes all the
important variables used in this study. Table 3 shows the descriptive statistics of all the
variables and compares the mean and t-test of each variable between the diversified and the
non-diversified firms. Table 4 shows the correlation matrix for the important variables.
Table 3 shows the descriptive statistics of all the important variables used in this study.
In section (1), important descriptive statistics like mean, median, standard deviation,
minimum value, maximum value and number of observations of the whole sample study are
Suan Poh and Chee-Wooi Hooy
24
Table 2: Description of variables
Variable Name Variable Description
Dependent Variables
Local Delay
(DELAYL)
Stock price delay relative to local market information, which is constructed
using equation (7)
Global Delay
(DELAYG)
Stock price delay relative to global market information, which is constructed
using equation (8)
Independent Variables
Foreign Sales
Dummy
(DIVERSE)
Dummy variable, which is equal to 1 if the company has more than 10% sales
in foreign countries
Number Foreign
Countries
(FCOUNTRY)
Number of foreign countries that a firm diversifies in
Foreign Sales Ratio
(FSALES)
Ratio of foreign sales to total sales
Herfindahl Index
(HERFINDAHL)
Herfindahl Index is constructed using equation (10)
Control Variables
Firm Size
(lnMCAP)
Natural logarithm of market capital of equity in millions of US dollars
Trading Volume
(VOLUME)
Average number of shares traded monthly scaled by total shares outstanding
Volatility
(VOLATILITY)
Standard deviation of weekly returns for a year
Number of
Analysts
(ANALYST)
Number of analyst forecasts that is reported by the Institutional Brokers‟
Estimate System
displayed. In sections (2) and (3), the descriptive statistics of the variables are further
divided according to whether the firm is diversified or non-diversified. The final section
shows the t-test results regarding the difference in the mean between the diversified and
non-diversified firms for all the variables.
From section (1), the delay in the stock price relative to the local index (DELAYL) has
a mean value of 0.413, which is higher than 0.218, the delay in the stock price relative to
global index (DELAYG). This shows that the stock price generally reacts slower to the local
index compared to the global index. For the independent variables, all four geographical
diversification proxies show a median of zero, which indicates that more than half of the
companies in our sample study are non-diversified. Firm size (MCAP) has a large different
value between the mean (US$1238m) and median (US$111m), which indicates that the
tabulation of firm size is skewed to the right with the majority of firms having a small firm
size and a small group of firms having a very large firm size. The monthly average trading
volume (VOLUME) also shows a large deviation between the mean (0.133) and the median
(0.038), which indicates that although most of the stocks are traded normally there is a small
group of stocks that are actively traded. The number of analysts (ANALYST) has a median
of zero, which shows that more than half of the stocks are not covered by analysts reports by
the Institutional Brokers‟ Estimate System. In sections (2) and (3), diversified firms
generally have a smaller delay than non-diversified firms irrespective of whether relative to
local or global information at the 1% significance level from the t-test. The relationship
between diversification and price delay is further tested using the regression model. All four
Suan Poh and Chee-Wooi Hooy
26
geographical diversification proxies show consistent descriptive statistics since the number
of foreign countries, foreign sales ratio and Herfindahl index show a higher mean value for
diversified firms compared to non-diversified firms with a t-test significance level of 1%.
Diversified firms record a firm size mean value of US$2066m, which is higher than the
US$1132m for non-diversified firms. This is reasonable because multinational companies
usually have a larger firm size as their business segments have expanded to other foreign
countries all over the world. The difference between the monthly average trading volume
between diversified and non-diversified firms is negligible and does not pose a significant t-
test result. The volatility of the weekly stock return for diversified firms is lower than that
for non-diversified firms because diversified firms can reduce their unsystematic risks
through diversification to other foreign countries. Lastly, there is more analysts‟ coverage
on diversified firms compared to non-diversified firms, which shows the preference of
analysts towards diversified firms.
3.2.1 OLS, Fixed Effects and Random Effects
Three different regression models – ordinary least squares model, fixed effects model and
random effects model – are run for all regression models, which include the regression
model with control variables only and the regression model with different types of
geographical diversification proxies including foreign sales dummy, number of foreign
countries, foreign sales ratio and Herfindahl Index. Most of the results show an inconsistent
sign and a significant level among the coefficient of variables. For example, DIVERSE. The
FCOUNTRY and HERFINDAHL are significant at 1% in all three models but have
different signs. As a result, it is appropriate to determine which model is the most suitable.
The results of all the regression models are not shown in this table due to space constraints.
3.2.2 Breusch and Pagan Lagrangian Multiplier Test
The Breusch and Pagan Lagrangian multiplier tests are used to identify whether normal
OLS or the random effects model is a better choice for the regression. If the significance
level is below 10%, it suggests that the random effects model should be chosen over normal
OLS. All regression models attain a significance level below 1%, which means that the
random effects model is preferable to the normal OLS for all models.
Table 5: Summary results of Breusch and Pagan Lagrangian Multiplier tests
Chi2 Control
model
Foreign
Sales
Dummy
Model
Number of
Foreign Countries
Model
Foreign
Sales Ratio
Model
Herfindahl
Index Model
Local Delay 114.79*** 120.35*** 114.72*** 121.02*** 115.42***
Global Delay 8663.76*** 4593.97*** 4595.50*** 4591.25*** 4593.37***
. ***, ** and * denote statistical significance at the 1%, 5% and 10% levels, respectively.
3.2.3 Hausman Test
Hausman test is used to identify whether the random effects model or the fixed effects
model is a better choice for the regression. If the significance level is below 10%, it
suggests that the fixed effects model should be chosen over the random effects model. All
the regression models have a significance level below 1%, which means that the fixed
effects model is preferable to the random effects model for all models.
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
27
Table 6: Summary results of Hausman test
Chi2 Control
model
Foreign Sales
Dummy
Model
Number of
Foreign
Countries Model
Foreign Sales
Ratio Model
Herfindahl
Index Model
Local Delay 363.90*** 328.77*** 341.39*** 313.45*** 337.16***
Global
Delay
148.62*** 114.64*** 115.84*** 116.12*** 115.74***
. ***, ** and * denote statistical significance at the 1%, 5% and 10% levels, respectively.
3.2.4 Fixed Effects Model
After running the Breusch and Pagan Lagrangian multiplier tests and the Hausman test, the
results suggest that the fixed effects model is the most suitable for our study for both local
delay and global delay. Table 7 and Table 8 show the fixed effects model for four different
geographical diversification proxies for both local and global delay.
Table 7 shows the fixed effects models for local delay with different geographical
diversification proxies. All the geographical diversification proxies show a significant and
positive effect on local delay, where DIVERSE, FCOUNTRY and HERFINDAHL show
significant at 1% and FSALES show significant at 5%. The results shown are consistent
with our hypothesis, which postulates that geographical diversification will increase the
local delay since investors do not favour companies that have a complex business structure.
When a company undergoes geographical diversification, its business coverage area
becomes larger and the company is exposed to other countries‟ risk in which its business is
involved. Its business structure becomes more complex than that of a company that only
focuses its sales locally (Morck and Yeung 1991). Since individual investors and
institutional investors prefer information that is easy to understand and widely available
(Chen 2005), many investors will abandon their research on diversified companies, which
will cause their potential investor base to become small and eventually decrease their
informational efficiency to local news. Due to the cost of information (Shapiro 2002), when
the information on a company or stock is hard to acquire or analyse, its informational
efficiency related to local news will decrease since most of the institutional investors are
more concerned about local news.
For the control variables, firm size (lnMCAP) shows a negative and significant effect
on local delay, which is consistent with previous literature (Bae et al. 2012). Firm size is an
important and prevalent determinant of delay, which is often used in previous literature,
such as Lim and Hooy (2010), Phillips (2010), Saffi and Sigurdson (2011) and Hou and
Moskowitz (2005). Larger firms usually have higher visibility and therefore less delay. The
monthly average trading volume (VOLUME) shows a positive and significant effect on
local delay. Stock return volatility shows a negative and significant result on local delay.
This result is consistent with previous literature since stocks with higher volatility will
attract more investors and therefore have less delay (Bae et al. 2012). It is surprising that the
number of analyst forecasts (ANALYST) shows a positive and significant result, since more
analyst forecasts should mean higher visibility of the firms.
We postulate that the positive result may be due to investors being strongly affected by
the forecast made by analysts until it causes a stock price delay to the local index. For
example, when the stock market crashes, many investors still believe in the analysts‟
forecast of a particular share and hold on to it; such a situation will cause the stock price to
experience a price delay relative to the local index.
Table 8 shows the fixed effects models for global delay with different geographical
diversification proxies. All the geographical diversification proxies – DIVERSE,
FCOUNTRY, FSALES and HERFINDAHL – show an insignificant effect on global delay.
Suan Poh and Chee-Wooi Hooy
28
Table 7: Fixed effects model for local delay This table shows the fixed effects model for local delay with the control variables and different geographical
diversification proxies. Column (1) only includes the control variables, while column (2), (3), (4), (5)
includes the control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and Herfindahl Index. The descriptions of all the
variables in this table are provided in Table 2. The regressions in this table are based on the following
equation:
(1) (2) (3) (4) (5)
CONSTANT 0.4301*** 0.4915*** 0.4918*** 0.4926*** 0.4904***
(0.0076) (0.0119) (0.0119) (0.0119) (0.0119)
lnMCAP -0.0002 -0.0074*** -0.0077*** -0.0074*** -0.0076***
(0.0014) (0.0021) (0.0021) (0.0021) (0.0021)
VOLUME 0.0088*** 0.0123*** 0.0123*** 0.0123*** 0.0123***
(0.0022) (0.0037) (0.0037) (0.0037) (0.0037)
VOLATILITY -0.4138*** -0.7170*** -0.7160*** -0.7188*** -0.7160***
(0.0288) (0.0408) (0.0408) (0.0408) (0.0408)
ANALYST 0.0045*** 0.0046*** 0.0046*** 0.0046*** 0.0045***
(0.0004) (0.0005) (0.0005) (0.0005) (0.0005)
DIVERSE 0.0160***
(0.0051)
FCOUNTRY 0.0073***
(0.0018)
FSALES 0.0281**
(0.0117)
HERFINDAHL 0.0524***
(0.0127)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.123 -0.188 -0.1877 -0.1882 -0.1877
R2 0.0082 0.0173 0.0175 0.0171 0.0175
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the statistical significance at the 1%, 5% and 10% levels, respectively.
The results show that whether a firm is geographical diversified or non-diversified does
not have any important effect on global delay. The different effects of the geographical
diversification proxies towards local delay and global delay suggest that there are two
different groups of investors who are monitoring the stock prices. The first group of
investors who take the local index into consideration when choosing stocks are strongly
affected by the geographical diversification decision of a firm. When a firm undergoes
geographical diversification, it will abandon the stock and will not make buy or sell
decisions on that stock, which causes the stock to become informationally inefficient. The
second group of investors who take global index into consideration when choosing stocks is
indifferent to whether or not a firm is geographically diversified.
For the control variables, firm size (lnMCAP) shows a negative and significant effect
on global delay, which is consistent with the result of the local delay model. The monthly
average trading volume (VOLUME) shows an insignificant effect on global delay. Stock
return volatility shows a positive and significant result on global delay. This result shows
that investors who emphasize global news do not favour stock with high volatility because
high volatility means high risk. The number of analyst forecasts (ANALYST) shows a
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
29
Table 8: Fixed effects model for global delay This table shows the fixed effects model for global delay with the control variables and different geographical
diversification proxies. Column (1) only includes the control variables, while columns (2), (3), (4), (5) include the
control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and Herfindahl Index. The descriptions of all the variables in this table are
provided in Table 2. The regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT 0.2517*** 0.2246*** 0.2240*** 0.2238*** 0.2240***
(0.0063) (0.0094) (0.0094) (0.0094) (0.0094)
lnMCAP -0.0064*** -0.0056*** -0.0057*** -0.0057*** -0.0057***
(0.0012) (0.0016) (0.0016) (0.0016) (0.0016)
VOLUME -0.0022 -0.0009 -0.0009 -0.0009 -0.0009
(0.0018) (0.0029) (0.0029) (0.0029) (0.0029)
VOLATILITY 0.2710*** 0.3128*** 0.3133*** 0.3132*** 0.3133***
(0.0241) (0.0322) (0.0322) (0.0322) (0.0322)
ANALYST -0.0015*** -0.0008** -0.0008** -0.0008** -0.0008**
(0.0004) (0.0004) (0.0004) (0.0004) (0.0004)
DIVERSE -0.0028
(0.0040)
FCOUNTRY 0.0004
(0.0014)
FSALES 0.0061
(0.0092)
HERFINDAHL 0.0018
(0.0101)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.1263 -0.202 -0.2021 -0.202 -0.2021
R2 0.0053 0.0057 0.0056 0.0056 0.0056
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the
statistical significance at the 1%, 5% and 10% levels, respectively.
negative and significant result, which is consistent with previous literature since more
analyst forecasts means higher visibility of the firms (Bae et al. 2012).
3.3 Robustness Test
The robustness test of this study can be divided into four sections. In the first section, the
white standard error is employed to solve the heteroscedasticity problems. VIF proxies are
used in the second section to show that our regression results do not face serious
multicollinearity problems. In the third section, we employ an alternative measure for both
local and global delay to show that our dependent variables are robust to other
measurements of delay. In the last section, we employ a crisis dummy to control for the
financial crisis effect.
3.3.1 White Standard Error
We solve the heteroscedasticity problem by using the White standard error, as suggested by
White (1980). The White standard error is used to solve the within cross-sectional
correlation and arbitrary heteroscedasticity issues. Table 9 shows the results of the local
delay for the fixed effects model using the White standard error. The results show that
although the standard error of our subject variables do increase (standard error of DIVERSE
Suan Poh and Chee-Wooi Hooy
30
Table 9: White standard error for local delay This table employs the White standard error on the fixed effects model of local delay, as shown in Table 7. Column
(1) only includes the control variables, while columns (2), (3), (4), (5) include the control variables and different
geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and the Herfindahl Index. The descriptions of all the variables in this table are provided in Table 2. The
regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT 0.4301*** 0.4915*** 0.4918*** 0.4926*** 0.4904***
(0.0084) (0.0136) (0.0136) (0.0136) (0.0136)
lnMCAP -0.0002 -0.0074*** -0.0077*** -0.0074*** -0.0076***
(0.0015) (0.0024) (0.0024) (0.0024) (0.0024)
VOLUME 0.0088 0.0123** 0.0123** 0.0123** 0.0123**
(0.0061) (0.0058) (0.0058) (0.0059) (0.0058)
VOLATILITY -0.4138*** -0.7170*** -0.7160*** -0.7188*** -0.7160***
(0.0361) (0.0518) (0.0515) (0.0517) (0.0517)
ANALYST 0.0045*** 0.0046*** 0.0046*** 0.0046*** 0.0045***
(0.0004) (0.0005) (0.0005) (0.0005) (0.0005)
DIVERSE 0.0160***
(0.0054)
FCOUNTRY 0.0073***
(0.0019)
FSALES 0.0281**
(0.0128)
HERFINDAHL 0.0524***
(0.0135)
N 49626 29339 29339 29339 29339
Adjusted R2 0.0081 0.0171 0.0173 0.0169 0.0174
R2 0.0082 0.0173 0.0175 0.0171 0.0175
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the
statistical significance at the 1%, 5% and 10% levels, respectively.
increases from 0.0051 to 0.0054, standard error of FCOUNTRY increases from 0.0018 to
0.0019, standard error of FSALES increases from 0.0117 to 0.0128, standard error of
HERFINDAHL increases from 0.0127 to 0.0135), the significant level remains unchanged.
As a result, we can conclude that the result remains significant after accounting for the
heteroscedasticity problem.
Table 10 shows the results of the global delay for the fixed effects model using the
White standard error to solve the heteroscedasticity problem. The results show that all the
geographical diversification proxies remain insignificant.
3.3.2 Multicollinearity Problem
We use the VIF indicator to show whether our regression models have multicollinearity
problems. If the VIF indicator shows a value of more than 5 or 10, it shows that there is a
serious multicollinearity problem in the regression model (O‟Brien 2007). The VIF proxies
for all the variables used in all the models have a value of less than 5, with the highest being
1.45, which is lnMCAP. The results show that there are no serious multicollinearity
problems present in our regression models.
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
31
Table 10: White standard error for global delay This table employs the White standard error on the fixed effects model of global delay, as shown in Table 8.
Column (1) only includes the control variables, while columns (2), (3), (4), (5) include the control variables and the
different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and the Herfindahl Index. The descriptions of all the variables in this table are provided in Table 2. The
regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT 0.2517*** 0.2246*** 0.2240*** 0.2238*** 0.2240***
(0.0070) (0.0104) (0.0103) (0.0104) (0.0104)
lnMCAP -0.0064*** -0.0056*** -0.0057*** -0.0057*** -0.0057***
(0.0012) (0.0018) (0.0018) (0.0018) (0.0018)
VOLUME -0.0022* -0.0009 -0.0009 -0.0009 -0.0009
(0.0012) (0.0021) (0.0021) (0.0021) (0.0021)
VOLATILITY 0.2710*** 0.3128*** 0.3133*** 0.3132*** 0.3133***
(0.0289) (0.0415) (0.0416) (0.0416) (0.0415)
ANALYST -0.0015*** -0.0008** -0.0008** -0.0008** -0.0008**
(0.0003) (0.0004) (0.0004) (0.0004) (0.0004)
DIVERSE -0.0028
(0.0042)
FCOUNTRY 0.0004
(0.0016)
FSALES 0.0061
(0.0106)
HERFINDAHL 0.0018
(0.0109)
N 49626 29339 29339 29339 29339
Adjusted R2 0.0052 0.0055 0.0055 0.0055 0.0055
R2 0.0053 0.0057 0.0056 0.0056 0.0056
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the statistical significance at the 1%, 5% and 10% levels, respectively.
Table 11: Summary results of VIF proxies VIF Control Model Foreign Sales
Dummy Model
Number of Foreign
Countries Model
Foreign Sales
Ratio Model
Herfindahl
Index Model
lnMCAP 1.42 1.44 1.45 1.44 1.44
VOLUME 1.01 1.01 1.01 1.01 1.01
VOLATILITY 1.1 1.09 1.09 1.09 1.09 ANALYST 1.32 1.34 1.34 1.34 1.34
DIVERSE 1.02
FCOUNTRY 1.02 FSALES 1.01
HERFINDAHL 1.02
Mean 1.21 1.18 1.18 1.18 1.18
3.3.3 Alternative Measure for Delay
Hou (2007) suggested that an alternative measure of delay can be used to replace the
original delay measure, which is calculated using equations (4) and (5). The alternative price
delay measure is indicated in the following equation:
( ) (9)
Suan Poh and Chee-Wooi Hooy
32
According to Hou (2007), there are a few engaging properties on this transformed version of
price delay measure. First, it is monotonic in x. Second, the logistics transformation of price
delay helps to remove the excess skewness and kurtosis of the original price delay measure.
Finally, the values of this transformed delay measure are not being restricted within the
intervals [0.1].
Table 12 shows the local delay fixed effects regression model using the alternative
delay measure constructed using equation (9). The results show that all the geographical
diversification proxies remain significant and that the coefficients of all the variables remain
unchanged.
Table 13 shows the global delay fixed effects regression model by using the alternative
delay measure constructed using equation (9). The results show that all the geographical
diversification proxies remain insignificant and the coefficients of the all the variables
remain unchanged unless VOLUME becomes significant at 5%.
In conclusion, we suggest that by using the alternative delay measure, most of the
coefficient signs and significant levels of control variables do not change. The significant
level and coefficient signs of all the geographical diversification proxies remain the same.
3.3.4 Controlling for Crisis
From the descriptive statistics across years for the control variables (Figure 8, 10, 12), the
results show that there is a huge fluctuation during Year 2008 and Year 2009. We postulate
that the fluctuation is due to the sub-prime financial crisis, which occurred during Year 2008
and Year 2009. To control for the financial crisis, we add a crisis dummy variable for all the
observations in Year 2008 and Year 2009.
Table 14 shows the local delay fixed effects regression model after including the crisis
dummy for Year 2008 and Year 2009. The results of all the geographical diversification
proxies remain unchanged and significant. CRISIS has a negative effect on DELAYL
because during a crisis, investors are more alert to the local news and local index. Their
decision to buy and sell shares will highly depend on the local index.
Table 15 shows the global delay fixed effects regression model after including the
crisis dummy for Year 2008 and Year 2009. The results of all the geographical
diversification proxies. remain insignificant. CRISIS has a negative effect on DELAYG
because during a crisis, investors are more alert to the global news and global index. Their
decision to buy and sell shares will highly depend on the global index.
4. Conclusion
All four geographical diversification proxies – foreign sales dummy, number of foreign
countries, foreign sales ratio and the Herfindahl Index – show a positive and significant
effect on local delay but an insignificant effect on global delay. Firm size, volatility and
number of analyst forecasts show significant effects on both local and global delay whereas
trading volume only shows a significant effect on local delay. After running the Breusch and
Pagan Lagrangian multiplier tests and the Hausman test, the results suggest that the fixed
effects model is most suitable compared to the OLS model and random effects model. All
the regression models in this study are robust to heteroscedasticity and multicollinearity
problems. All geographical diversification proxies remain significant towards local delay
when we use alternative delay measures and control for financial crisis. The t-tests
regarding the difference in the mean between diversified and non-diversified firms for all
the variables show significant results except for trading volume.
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
33
Table 12: Alternative measure for local delay This table employs the alternative measure of local delay generated using equation (9) for the fixed effects model
shown in Table 7. Column (1) only includes the control variables, while columns (2), (3), (4), (5) include the
control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and Herfindahl Index. The descriptions of all the variables in this table are
provided in Table 2. The regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT -0.3710*** -0.0642 -0.0626 -0.0582 -0.0686
(0.0377) (0.0585) (0.0584) (0.0584) (0.0585)
lnMCAP 0.0033 -0.0320*** -0.0334*** -0.0319*** -0.0329***
(0.0069) (0.0103) (0.0103) (0.0103) (0.0103)
VOLUME 0.0378*** 0.0612*** 0.0612*** 0.0613*** 0.0612***
(0.0108) (0.0181) (0.0181) (0.0181) (0.0181)
VOLATILITY -1.8852*** -3.4399*** -3.4349*** -3.4487*** -3.4357***
(0.1431) (0.2012) (0.2012) (0.2012) (0.2012)
ANALYST 0.0212*** 0.0215*** 0.0214*** 0.0215*** 0.0214***
(0.0022) (0.0025) (0.0025) (0.0025) (0.0025)
DIVERSE 0.0769***
(0.0250)
FCOUNTRY 0.0354***
(0.0090)
FSALES 0.1278**
(0.0576)
HERFINDAHL 0.2396***
(0.0628)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.1242 -0.1891 -0.1888 -0.1893 -0.1889
R2 0.0072 0.0163 0.0166 0.0162 0.0165
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the
statistical significance at the 1%, 5% and 10% levels, respectively.
Suan Poh and Chee-Wooi Hooy
34
Table 13: Alternative measure for global delay This table employs the alternative measure for global delay generated using equation (9) for the fixed effects
model, as shown in Table 8. Column (1) only includes the control variables, while columns (2), (3), (4), (5) include
the control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and the Herfindahl Index. The descriptions of all the variables in this table are
provided in Table 2. The regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT -1.3515*** -1.5477*** -1.5529*** -1.5547*** -1.5527***
(0.0411) (0.0650) (0.0650) (0.0650) (0.0651)
lnMCAP -0.0419*** -0.0311*** -0.0317*** -0.0321*** -0.0317***
(0.0075) (0.0114) (0.0114) (0.0114) (0.0114)
VOLUME -0.0245** -0.0469** -0.0470** -0.0470** -0.0470**
(0.0118) (0.0201) (0.0201) (0.0201) (0.0201)
VOLATILITY 1.2593*** 1.4748*** 1.4786*** 1.4788*** 1.4785***
(0.1563) (0.2238) (0.2238) (0.2238) (0.2238)
ANALYST -0.0120*** -0.0089*** -0.0090*** -0.0091*** -0.0090***
(0.0024) (0.0028) (0.0028) (0.0028) (0.0028)
DIVERSE -0.0332
(0.0278)
FCOUNTRY -0.0004
(0.0100)
FSALES 0.0345
(0.0641)
HERFINDAHL -0.0039
(0.0699)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.1277 -0.2045 -0.2046 -0.2046 -0.2046
R2 0.004 0.0036 0.0035 0.0036 0.0035
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the statistical significance at the 1%, 5% and 10% levels, respectively
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
35
Table 14: Crisis controlling dummy for local delay This table adds in a crisis dummy for all the observations in Year 2008 and Year 2009 for the fixed effects model
shown in Table 7. Column (1) only includes the control variables, while columns (2), (3), (4), (5) include the
control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and the Herfindahl Index. The descriptions of all the variables in this table are
provided in Table 3. The regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT 0.4363*** 0.5098*** 0.5101*** 0.5109*** 0.5087***
(0.0076) (0.0119) (0.0119) (0.0119) (0.0119)
lnMCAP -0.0016 -0.0115*** -0.0117*** -0.0115*** -0.0117***
(0.0014) (0.0021) (0.0021) (0.0021) (0.0021)
VOLUME 0.0091*** 0.0119*** 0.0119*** 0.0119*** 0.0118***
(0.0022) (0.0037) (0.0037) (0.0037) (0.0037)
VOLATILITY -0.3455*** -0.5501*** -0.5496*** -0.5515*** -0.5491***
(0.0300) (0.0429) (0.0429) (0.0429) (0.0429)
ANALYST 0.0045*** 0.0046*** 0.0046*** 0.0046*** 0.0046***
(0.0004) (0.0005) (0.0005) (0.0005) (0.0005)
CRISIS -0.0181*** -0.0379*** -0.0378*** -0.0380*** -0.0379***
(0.0022) (0.0030) (0.0030) (0.0030) (0.0030)
DIVERSE 0.0159***
(0.0051)
FCOUNTRY 0.0070***
(0.0018)
FSALES 0.0293**
(0.0117)
HERFINDAHL 0.0521***
(0.0127)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.1214 -0.1805 -0.1803 -0.1807 -0.1802
R2 0.0096 0.0235 0.0237 0.0233 0.0237
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the statistical
significance at the 1%, 5% and 10% levels, respectively
Suan Poh and Chee-Wooi Hooy
36
Table 15: Crisis controlling dummy for global delay This table adds in a crisis dummy for all the observations in Year 2008 and Year 2009 for the fixed effects model
shown in Table 8. Column (1) only includes the control variables, while columns (2), (3), (4), (5) include the
control variables and different geographical diversification proxies, which are foreign sales dummy, number of foreign countries, foreign sales ratio and the Herfindahl Index. The descriptions of all the variables in this table are
provided in Table 3. The regressions in this table are based on the following equation:
(1) (2) (3) (4) (5)
CONSTANT 0.2710*** 0.2429*** 0.2424*** 0.2421*** 0.2424***
(0.0063) (0.0094) (0.0094) (0.0094) (0.0094)
lnMCAP -0.0108*** -0.0097*** -0.0097*** -0.0098*** -0.0097***
(0.0012) (0.0017) (0.0017) (0.0017) (0.0017)
VOLUME -0.0013 -0.0013 -0.0013 -0.0013 -0.0013
(0.0018) (0.0029) (0.0029) (0.0029) (0.0029)
VOLATILITY 0.4862*** 0.4803*** 0.4807*** 0.4807*** 0.4807***
(0.0248) (0.0338) (0.0338) (0.0338) (0.0338)
ANALYST -0.0015*** -0.0008** -0.0008** -0.0008** -0.0008**
(0.0004) (0.0004) (0.0004) (0.0004) (0.0004)
CRISIS -0.0570*** -0.0380*** -0.0380*** -0.0380*** -0.0380***
(0.0019) (0.0024) (0.0024) (0.0024) (0.0024)
DIVERSE -0.0029
(0.0040)
FCOUNTRY 0.0002
(0.0014)
FSALES 0.0073
(0.0092)
HERFINDAHL 0.0015
(0.0100)
N 49626 29339 29339 29339 29339
Adjusted R2 -0.1025 -0.1898 -0.1898 -0.1898 -0.1898
R2 0.0263 0.0158 0.0158 0.0158 0.0158
Notes: The standard error of each coefficient is reported in parentheses. N denotes the number of observations
Adjusted R2 represents the adjusted R-squared, while R2 represents R-squared. ***, ** and * denote the
statistical significance at the 1%, 5% and 10% levels, respectively.
References Bae, K. H., Ozoguz, A., Tan, H. and Wirjanto, T. S. 2012. Do foreigners facilitate information
transmission in emerging markets? Journal of Financial Economic 105(1): 209–227.
Barney, J.B. and Hesterly, W.S., 2008. Strategic management and competitive advantage: Concepts
and cases. Prentice Hall, Upper Saddle River, NJ.
Bartlett, C. A. and Ghoshal, S. 2000. Transnational management: Text, cases and readings in cross-
border management. McGraw-Hill.
Bodnar, G.M., Tang, C. and Weintrop, J. 1999. Both sides of corporate diversification: The value
impacts of geographic and industrial diversification. NBER Workingpaper 6224.
Breusch, T. S., and Pagan A.R. 1980. The Lagrange Multiplier Test and its applications to model
specification in econometrics. Review of Economic Studies 47(1): 239-253.
Effect of Geographical Diversification on Informational Efficiency in Emerging Countries
37
Callen, J. L., Khan, M., and Lu, H. 2013. Accounting quality, stock price delay, and future stock
returns. Contemporary Accounting Research 30(1): 269–295.
Campbell, J.Y., Lo, A.W., and MacKinlay, A.C. 1997. The econometrics of financial markets.
Princeton: Princeton University Press.
Chang, S.C., and Wang, C.F., 2007. The effect of product diversification strategies onthe relationship
between international diversification and firm performance. Journal of World Business 42 (1):
61–79.
Chen, J. 2005. Information theory and market behavior. ICFAI Journal of Behavioral Finance 2(4):
25-45.
Chordia, T., and Swaminathan, B. 2004. Incomplete information, trading costs and cross-
autocorrelations in stock returns. Economic Notes, 33(1): 145–181.
Denis, D.J., Denis, D.K and Yost, K. 2002. Global diversification, industrial diversification, and firm
value. Journal of Finance 57 (5), 1951–1979.
Fama, E.F. 1965. The behavior of stock-market prices. Journal of Business 38: 34-105.
Ghoshal, S. (1987). Global strategy: An organizing framework. Strategic Management Journal 8(5):
425–440.
Granger, C.W.J., and Morgenstern, O. 1963. Spectral analysis of New York stock market prices.
Kyklos, 16, 1-27.
Hitt, M.A., Hoskisson, R.E., and Kim, H. 1997. International diversification: effects on innovation and
firm performance in product-diversified firms. Academy of Management Journal 40 (4), 767–
798.
Hou, K., and Moskowitz, T. J. 2005. Market frictions, price delay, and the cross-section of expected
returns. Review of Financial Studies, 18(3): 981–1020.
Kogut, B. 1985. Designing global strategies: Comparative and competitive value-added chains. Sloan
Management Review,26(4): 15.
Kogut, B., and Kulatilaka, N. 1994. Operating flexibility, global manufacturing, and the option value
of a multinational network. Management Science 40(1): 123–139.
Lesmond, D. A. 2005. Liquidity of emerging markets. Journal of Financial Economics 77(2): 411–
452.
Lim, K.P., and Hooy, C.W. 2010. The delay of stock price adjustment to information: A country-level
analysis. Economics Bulletin 30: 1609-1616
Lim, K.P., and Hooy, C.W. 2013. Foreign shareholding and stock price efficiency: Firm-level
evidence from Malaysia. SSRN Working Paper.
Lo, A.W., and MacKinlay, A.C. 1988. Stock market prices do not follow random walks: Evidence
from a simple specification test. Review of Financial Studies 1: 41-66.
Lu, J. W., and Beamish, P. W. 2004. International diversification and firm performance: The s-curve
hypothesis. Academy of Management Journal 47(4): 598–609.
Merton, R. C. 1987. A simple model of capital market equilibrium with incomplete information,
Journal of Finance, 42: 483-510.
Morck, R., and Bernard Y. 1991. Why investors value multinationality, Journal of Business 64: 165–
187.
Pavelin, S and Barry, F. 2005. The single market and the geographical diversification of leading firms
in the EU. The Economic and Social Review 36(1): 1–17.
Phillips, B. 2011. Options, short-sale constraints and market efficiency: A new perspective. Journal of
Banking and Finance 35(2): 430–442.
Porter, M. E. 1986. Changing patterns of international competition. California Management Review
28(2): 9–40.
Saffi, P. A. C., and Sigurdsson, K. 2011. Price efficiency and short selling. Review of Financial
Studies 24(3): 821–852.
Tallman, S., and Li, J. 1996. Effects of international diversity and product diversity on the
performance of multinational firms. Academy of Management Journal 39(1): 179.
White, H. 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity. Econometrica 48(4): 817-838.